[Qgis-developer] Qgis-developer Digest, Vol 138, Issue 36

[Andreas Plesch]

> Date: Tue, 11 Apr 2017 04:41:32 -0700 (MST)
> From: Tom Chadwin <tom.chadwin at nnpa.org.uk>
>
> Yes, I mean that all models are made up of discrete algorithms, and those
> algorithms will only really work as intended if chained together as in the
> model. Nevertheless, the algorithms will appear in the Processing toolbox,
> so users could try to use them in isolation, and have difficulty forming
> the
> required inputs. Or do you prevent the individual algorithms appearing in
> the Processing toolbox?
>
>
I believe there is always the option to turn a model into an algorithm
using the Export as Python script functionality of the model editor.

The underlying algorithms still need to be available but could be in their
own group ('advanced', 'atomic', 'dangerzone :)') and the integrated
algorithm could be named accordingly as well ('runme :)').

But in the end it comes down to making more universally (perhaps within the
problem domain) useful algorithms and providing good documentation ( which
is pretty easy using the built-in help system ). For example, I came to the
conclusion that is useful to have a very basic algorithm 'write string to
file' as a basic building block of models using algorithms which output
strings and digest files.

I had similar concerns and started to think that it should be possible to
'compile' a script (recursively) which uses processing.runalg down to a
script which only uses 'native' api functionality and 'native' algorithms.
Effectively, patching in the called script source (perhaps as a function?).
The 'compiler' could perhaps be an algorithm itself. This way it would be
easy to provide just a single algorithm which is not depending on other
custom algorithms.

-Andreas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20170411/04a1fee9/attachment.html>